I watched a discussion tonight from Nant Studios for a PGA/VES panel on using Unreal Engine as a narrative tool. The useful part was not “look, game engines can make movie images now.” That conversation is already tired.
The useful part was that the panel had people who are deep enough in the work to talk about where it helps, where it breaks, and what happens when a cool tool runs into a real schedule. Fulvio Sestito and Paul “Pizza” Pianezza from Realdream Studio, Eri Hawkins from Jim Henson Company, Emanuel Druckmann from Lightspeed LA, and Zach Staenberg, the Oscar-winning editor from The Matrix trilogy, were not selling a fantasy version of production. They were talking about dailies, lenses, mocap suits, facial rigs, LiDAR scans, line producers, ADs, budgets, post handoffs, and the old problem of getting everyone to understand the same shot before it costs a fortune.
That is the real reason game engines matter to Hollywood. Not because everything should look like a game. Not because virtual production makes every project smarter. Game engines matter because they shorten the distance between imagination and production reality. They let teams test the space before the space exists.
The strategic shift is not real-time graphics. The strategic shift is moving decisions upstream.
The engine is becoming a production room
The old version of this conversation was about previs. That is still part of it, but it is too small a frame now.
Real-time engines can help with previs, techvis, postvis, virtual scouting, LED volume work, in-camera VFX, set design review, camera blocking, lens choices, lighting planning, VFX planning, location planning, stage planning, and client approvals. They can also become the place where assets move between departments instead of getting rebuilt five different ways by five different vendors.
That does not make the engine magic. It makes it a coordination layer.
On a real job, the problem is rarely that nobody can imagine something cool. The problem is that the director’s idea, the agency deck, the production designer’s model, the DP’s plan, the VFX supervisor’s assumptions, the line producer’s budget, the client’s approval, and the post team’s delivery path are often living in separate places.
A shared real-time scene can expose the gaps earlier.
Can the camera physically be where the board says it is? Does that lens make sense in that space? Does the LED wall need a different background plate? Is the practical set tall enough? Is the extension doing more work than anyone budgeted for? Does the lighting direction match the location? Does the actor have somewhere to move? Does the VFX team need a scan before the stage is struck? Is the client approving scale, mood, or actual geography?
Those are not abstract questions. Those are the places where money disappears.
Postvis saves The Killer
Zach told the best practical story of the night. He was working on John Woo’s The Killer when the SAG strike shut down the actor side of the production. They were shooting in Paris. The actors had to leave, but the French crew was still contracted for more time, so the team shot what they could with stunt performers: wide shots, back-of-head material, action pieces that could still be useful.
That left Zach and Woo with a pile of footage and a lot of missing connective tissue.
Fulvio came into the editing suite with a mocap suit and built a couple blocks of Paris in Unreal based on the location. This was not a giant department disappearing for a month and coming back with a precious presentation. It was much closer to production problem-solving in the room. The director and editor could pick lenses, choose camera rigs, check what those rigs could actually do, look at the height off the floor, test a move, change the lens, adjust the blocking, and get material into Avid quickly enough to cut against the live-action footage.
In a traditional previs workflow, Zach said, you would give notes and get shots back a week later. Here, if they were pushing, they could get useful material the same day.
That time difference changes the conversation. You are not waiting for a render to find out whether the idea is dead. You are arguing with the shot while the people who can change it are still there.
One Universal exec looked at the cut with postvis mixed into live action and asked which was which. That does not mean the Unreal material was final. It means it was good enough, in context, to let people make editorial and production decisions instead of staring at a blank spot and guessing.
That is the point.
The planning tool can create planning problems
The same The Killer example also showed the trap.
Zach described the postvis becoming a kind of bible for what they wanted to shoot. That is powerful. It gave everyone a clear target. But because the workflow came together under pressure, some production voices saw the work late. The line producer and AD looked at shots and said, essentially, that one is too expensive for the budget.
That is not a failure of Unreal. That is a workflow warning.
If the engine is being used to make production decisions, production has to be in the room early enough to shape them. Otherwise, you can create beautifully clear impracticality. The director loves it, the editor can cut it, the client can understand it, and then the people responsible for the day have to explain why the shot cannot happen.
That is why this is not just a VFX toy. The more useful the engine becomes, the more it needs producer discipline around it.
Who is reviewing for budget? Who is reviewing for schedule? Who is checking rigging, lenses, locations, safety, crew time, LED constraints, art department implications, VFX handoff, and post delivery? What is accurate in the model, and what is only a storytelling placeholder?
The tool moves decisions earlier. It does not make those decisions automatically correct.
Digital puppetry, not just digital pictures
Eri’s work at Jim Henson Company pushed the conversation into a different category. They built a fully digital virtual production environment that works more like a three-camera sitcom. No live-action plate. No compositing in the normal sense. Motion capture performers handle bodies while Henson puppeteers perform faces. Virtual cameras sit on physical pedestals. Performers can see themselves as digital characters in the environment while they are performing.
That matters because puppetry is brutally sensitive to latency. If the feedback is late, the performance suffers. Earlier versions of the workflow had limits: low-resolution assets, flatter lighting, more meetings about what the final thing might look like. Unreal let them bring more of the creative conversation to the front of the process.
That phrase is worth holding onto: bring the creative conversation to the front.
Lighting was a real example. If you cannot see dramatic lighting during the performance, you hedge. You keep it flat because nobody wants the character to disappear into a shadow no one could judge on the day. Once the performers and the team can see the scene working in real time, lighting becomes part of the performance and direction instead of a downstream correction.
The newest Henson example was real-time dynamic interactive fur on a project called Monster Jam, with monsters dancing. The fur reacts to the performance in real time, so performers can see it and incorporate that motion back into the work. That is a small sentence with a big production implication: the simulation is not just a visual layer added later. It becomes part of what the performer reacts to.
That is where these tools get interesting. Not when they replace judgment. When they let departments and performers react to things that used to arrive too late.
Small teams, more options, more danger
Pizza talked about helping friends win a commercial pitch. The ask was basically an animatic. Realdream brought it into Unreal and used the same time to generate multiple shot ideas and approaches instead of one fixed version.
That is the obvious appeal. A smaller team can move through more options faster. You can test a camera idea, a lighting idea, a piece of action, or a location approach without waiting for a traditional render-review cycle. Emanuel made the same point from the game side: in a real-time scene, moving a light or moving a camera can become a two-minute conversation instead of a next-day review.
The good version of that is better exploration. The bad version is endless thrashing.
Eri said the best thing and the hardest thing about Unreal are the same thing: you can iterate pretty much endlessly. The question is when you make it stop.
That is a producer question.
Traditional pipelines have external brakes. You send work out. You wait. A render takes time. A vendor needs a brief. A review happens tomorrow. Those brakes are annoying, but they also force decisions. Real-time tools remove some of that friction, which means the team needs a different kind of discipline.
Sometimes the right answer is to make the image look worse on purpose. Pizza mentioned reviewing in grayscale so people do not get distracted by the wrong issue. If the question is blocking, do not let everyone argue about color. If the question is geography, do not let everyone chase wardrobe texture. If the question is whether the camera move works, label the set dressing as approximate and keep the room on task.
Real-time does not remove review structure. It makes review structure more important.
The Quarry and the asset question
The most obvious film/game boundary case was The Quarry, the choose-your-own-adventure horror game. They shot roughly 33 hours of performance capture with 13 actors. That is mini-series scale content living inside a game pipeline.
The director could look at dailies, evaluate performances, and place cameras with a level of freedom that does not map cleanly to a traditional shoot. Actors in mocap suits could see themselves as digital characters inside the actual game environments on monitors on the stage. The team could move the center of the stage around and keep working in different parts of the virtual set.
This is where producers should pay attention to asset reuse. A film asset, a game asset, a previs asset, a VFX asset, and a marketing asset are not automatically the same thing. They have different fidelity, topology, texture, rigging, lighting, file format, performance, and licensing requirements.
But the direction of travel is obvious: departments are going to want to reuse more of the same world. Not because reuse is always clean, but because rebuilding the same idea from scratch in every department is stupid when the schedule is tight and the client is already confused.
If the model is going to move across departments, someone has to own it. Someone has to maintain version control. Someone has to know which asset is reference, which is approved, which is final pixel, and which is a fast lie that helped the director make a decision on Wednesday.
That is not glamorous work. It is the work that keeps a promising workflow from becoming a folder full of mystery files.
When not to use it
The panel was refreshingly clear that not every project needs this.
A small dialogue film may not benefit from building a real-time production environment. A simple commercial may not need virtual scouting. A stage job may not need an LED volume. A beautifully built Unreal scene can still be the wrong answer if the budget, schedule, crew, and approval path do not justify it.
The technical limits are real too. Unreal handles some data well and fights other data. Getting material out into a traditional VFX pipeline can be painful. Passes and mattes are not always what compositors want. Depth of field is a post-process effect, not physically correct in the way a VFX pipeline may require. Axes and scale can bite you when moving between Maya and Unreal. Windows-centric workflows can run into the reality that many VFX pipelines are Linux-based.
There are bridges. USD matters. Larger studios can build tools. A team with TDs can solve problems a small run-and-gun team cannot. But the point stands: you have to know what job the engine is doing.
Is it previs? Techvis? Postvis? Final pixel? Reference? A virtual scout? A camera planning tool? A client approval tool? A live stage system? A source of assets for downstream VFX?
Those are different promises.
Why producers should care
Producers do not need to become game developers. They do need to understand what real-time engines change about timing, risk, approvals, and collaboration.
The value is not that the tool makes a cooler-looking temporary shot. The value is that ambiguity gets exposed earlier.
The director can see the blocking before shoot day. The DP can react to lens and lighting implications earlier. Art department can see scale in context. VFX can flag what needs scanning, tracking, extension, or replacement. Camera can test whether the move is physically plausible. Locations can understand what has to be real and what can be extended. Stage teams can look at LED volume implications before everyone is standing around waiting. Clients and agencies can react to something more concrete than a deck.
Better planning does not eliminate production problems. Nothing does. It moves some problems to cheaper moments.
That is enough reason to care.
If a $5,000 decision can prevent a $50,000 shoot-day surprise, that is not a toy. If a virtual scout can get the client to approve eyelines, scale, geography, and background direction before the build starts, that is not a toy. If postvis can help editorial, VFX, and production agree on what the missing shot needs to accomplish, that is not a toy.
It is production infrastructure.
Bad use, good use
Bad: Build a virtual production pipeline because it looks futuristic.
Good: Use the engine to answer whether the set, camera move, lighting plan, and VFX extension can actually work before shoot day.
Bad: Treat the Unreal scene as final truth.
Good: Label what is accurate, what is approximate, and what still needs verification.
Bad: Bring clients into a flashy virtual scout with no decision agenda.
Good: Use the scout to lock specific questions: eyelines, scale, background, lighting direction, action geography, and what must be built practically.
Bad: Make beautiful previs in isolation and hand it to production after the creative team is already in love with it.
Good: Bring the AD, line producer, art department, VFX, camera, and post into the review early enough to catch the expensive assumptions.
Bad: Assume asset reuse will happen because everything is “in Unreal.”
Good: Decide who owns the assets, what formats matter, what departments can trust them, and how versions get maintained.
The producer test
Before spending real money on the workflow, I would ask:
- What decision does the engine help us make earlier?
- Who owns the model and assets?
- Is this previs, techvis, final pixel, or reference?
- What departments need to trust it?
- What assumptions are baked into the scene?
- Does camera, lens, scale, and lighting match the real shoot?
- Can art department, VFX, camera, and post all use the same assets?
- What happens if the director changes the blocking?
- Is the workflow saving time, or just moving cost into a different department?
- What needs to be approved before shoot day?
- What can break on stage?
- Who maintains version control?
If the team cannot answer those questions, the engine may still be useful. But it is not yet a production plan.
The labels are behind the work
Someone asked whether a short film made in Unreal should be submitted to the Academy as animation or live action. The answer from the room was basically animation.
That sounds like a classification issue, but it points to the bigger thing happening. The work is moving faster than the categories. Game people are thinking more cinematically. Film people are using game engines to plan and sometimes make shots. Editors are cutting postvis that can fool an executive for a second. Puppeteers are performing digital characters with real-time fur. Directors are scouting virtual sets. VFX is asking whether the LiDAR scan can move into the engine. Producers are trying to figure out what is real, what is approved, and what is about to become expensive.
The labels will catch up eventually. The work will not wait.
Where this goes
Fulvio said people have been waiting for a couple of decades for games and film to meet in the middle. That is true, but I think the more useful producer version is narrower.
Real-time engines are becoming production infrastructure. They are not right for every job, and they do not excuse sloppy planning. They introduce their own costs, specialists, dependencies, versioning problems, and pipeline traps.
But the direction is clear. More production decisions are going to happen inside shared spatial models before the expensive part of the work begins. More clients will expect to see something they can react to earlier. More departments will expect the virtual version of the scene to answer practical questions. More assets will need to survive across previs, stage, VFX, editorial, marketing, and maybe interactive extensions.
Producers do not need to become game developers. But they do need to understand what these tools change: when decisions are made, who needs to be in the room, what assumptions are visible, what approvals mean, what can be reused, and where risk is hiding.
The question is not whether this workflow is worth learning. It is whether you will have enough time to catch up before it becomes table stakes.
Related reading
- What AI Is Actually Doing on Set
- Later reading: Before AI, It Was Just Graphics